Copilot Chat Assistant

Context

Live chat is an important lead source for car dealers.

As a leading chat platform and chat operations provider, Gubagoo supports hundreds of internal operators and external BDC teams.

The people handling these chats are responsible for managing many conversations at once, often completing repetitive tasks like collecting basic contact info or searching inventory.

Opportunity

Giving chat operators an AI assistant could help them focus on high-value conversations.

Today, our chat AI only handles simple customer conversations before handing them over to a real person, at which point, AI assistance is no longer available.

There are many situations AI could assist with if it was available to chat operators:

• autocomplete
• chat scripts
• suggesting vehicles
• summarizing conversations

Goal

Enable chat operators to focus on their most important conversations by delegating repetitive tasks to AI.

Select process and solutions

Risk: Chat operators were weary of AI

Letting AI take over an interaction with a customer can feel high-risk with the chance that it behaves unexpectedly and frustrates the customer, jeopardizing the prospect.

With this in mind, I drafted some design goals that should safeguard against these concerns:

Always understand what’s going on

I created simple labels to indicate the status of chats while Copilot is running. They indicate what Copilot is doing, or if user intervention is needed.

Minimizing cognitive load

Copilot suggests only the most relevant action, as to not overwhelm users. All additional actions are available via a dropdown in the chat actions bar.

Skill: Smart suggest vehicles based on customer inquiries

When a vehicle of interest is detected, Copilot suggests the best match based on inventory. The inventory search panel also auto-filters to match Copilots query in case the operator wants to hand-select a vehicle.

Skill: Autocomplete

Shows autocomplete suggestions for common phrases and saved responses.

Keeping users in control with safeguards

Once a user initiates a skill, they can monitor the progress using the status bars displayed on the chat cards. Because users can keep an eye on the status of Copilot, they should feel in control, and have greater trust in the system.

If anything goes wrong, Copilot will stop and alert the user to jump back into the conversation, minimizing customer frustration.

Easy reference

Copilot interactions are clearly marked in the transcript for accurate reference.

Results and impact

MVP of this feature is set to release Q1 2023, and we have more planned for future enhancements.

Our team is monitoring these metrics to gauge success of this feature: Number of chats handled by each operator, average chat queue time, average response time, chat duration.

In early testing with select chat operator teams, we’ve received positive feedback so far:

“This feature has allowed our chat operators to engage in more situations that require human interactions and has improved our response time all around. The AI handles all of the monotonous situations, which has improved our chat operator moral. We are hitting more team goals than ever thanks to CoPilot!” -Chat Operations Manager

“CoPilot makes my job so much easier and faster! …I feel like I have my own personal assistant all day long!” -Chat Operator

‍